systemprompt.io is the only AI infrastructure you actually own. One self-hosted binary that runs your MCP servers, governs every agent, and logs every action. Install it, build on it, keep it forever.
Governs Claude, OpenAI, Gemini, and Groq through one control plane
A single compiled Rust binary, PostgreSQL its only dependency. It runs your MCP servers, governs every agent through a synchronous, extensible pipeline, and writes SIEM-compatible JSON logs. Read the source. Extend it in code you own.
Host MCP servers and agents behind one authenticated gateway, on infrastructure you control.
Every model call, tool invocation, and agent decision flows through a pipeline you configure and can audit.
It is a library, not a black box. Extend it with Rust extensions, custom jobs, and your own MCP tools.
Three ways to govern AI. Two of them leave you dependent on someone else.
A synchronous governance pipeline and SIEM-compatible logs give your auditor evidence, not screenshots. When they ask what your AI did and who approved it, you have the answer.
Your systems, your data, your control. The software is yours to keep and inspect. Nothing leaves your business unless you allow it.
Self-hosted and fully ownedRuns entirely on your own systems with no outside connections required. Plug in your existing staff login and it handles the rest.
One system, no outside connectionsPasswords and keys are kept locked away and never shown to the AI. The AI asks for an action; the credential is added safely behind the scenes.
Credentials never exposed to AIEvery AI action is linked to the person who triggered it, timestamped, and stored as evidence. A complete trail from request to result.
A per-action record tied to a real personKey moments in AI activity can trigger alerts and feed the monitoring and reporting tools you already run.
Live alerts across AI activityTools and data are limited by role and team. Finance sees finance tools, engineering sees engineering tools, down to what is even visible.
Access limited by role and teamBuilt to support SOC 2, ISO 27001, and HIPAA programmes. Informed by the OWASP Top 10 for Agentic Applications.
Owning software nobody uses is worthless. We deliver the system you own, the process to run it, and the adoption to make it how your company actually works.
AI only matters when the people who do the work actually use it. We work alongside your teams until AI is part of how they get the job done every day, not a tool that gathers dust after launch.
Hands-on training turns raw capability into repeatable process. We teach your people how to use the system and how to build AI into the way work actually gets done, whether remote or in person from Valencia, Spain.
Underneath it all is the system: one piece of software you install on your own infrastructure that wraps, owns, and binds every AI in your business. Any provider, every tool, fully under your control.
The AI landscape changes monthly. Keeping a home-built control layer working means a permanent team chasing those changes. We absorb that work so your team doesn't have to. You could build it yourself. But by the time you ship it, you will need to rebuild it.
You own the result. You don't own the ongoing maintenance.
Almost nothing. The software is source-available under the Business Source License 1.1, it runs on your own infrastructure, and everything your teams build on it (tools, Skills, settings, records) is your property. There is no cloud account to switch off and no data held hostage.
Nothing walks out the door. The system, the data, and the work stay yours.
It runs entirely on your own systems, even with no internet connection at all. Nothing is sent outside unless you choose to. Your data stays in your business and in your country.
No data leaves your network. Your systems, your rules, your jurisdiction.
You set the rules once, centrally. Every AI action is checked against them automatically, and anything that breaks a rule is stopped before it happens. The same rules apply to every team and every tool.
One rulebook for the whole company, enforced automatically.
Every AI action records its cost. Group it by team, project, or provider to see exactly where the money goes, and compare providers side by side.
"What did we spend on AI last month, by team?" is a single question.
Enterprise AI governance and Claude Code resources backed by real production experience.
Enforcement lands 2 August 2026. The Article 9, 12, and 14 technical controls to run self-hosted, and the audit evidence you must produce.
Read guide
The AI management system standard explained: its clauses, the 38 Annex A controls, the certification path, and how it maps to NIST AI RMF.
Read guide
Charge agents per tool call with HTTP 402: the x402 payment flow, how payment layers with MCP auth, what agents pay for, and a full Rust implementation.
Read guideYou do. The software is source-available under the Business Source License 1.1, and you can clone the template repository on GitHub to inspect it before you commit. Everything your teams build (tools, skills, settings, and customisations) is your property. You only license the underlying software itself.
Nowhere you don't control. systemprompt.io runs on your own systems. Your data flows only to the AI providers you choose, and nothing leaves your network unless you set it up that way.
All the major ones: Claude, OpenAI, Google, plus your own private AI. You can switch whenever you like. It is a setting change, not a rebuild, and your rules and records carry over.
The AI tools your teams already use, such as Claude Code, Cursor, and Claude Cowork, plus any tools you build in-house. They all follow the same rules and feed the same records.
Every AI action can be sent straight to the security and reporting tools you already run. Everything is also stored in one place you can search directly.
Yes. Rules are set centrally and can differ by role and team. Engineering sees engineering tools, finance sees finance tools, and the rules are enforced automatically and consistently.
One binary you install on your own systems, with PostgreSQL as the database. No complex cloud setup, and it can run with no internet connection at all. Most teams go from trial to live in days, not months.
Free to evaluate, with no time limit. Using it in production needs a commercial licence. Book a call for a quote scoped to your business.
Three layers, delivered as one. We stand up the system inside your infrastructure, run hands-on training for your people, then work alongside them to embed AI into daily work until it is a habit and the processes are yours to run.
Remote, or in person from Valencia, Spain. Either way the training is hands-on and built around your teams and how they actually work.
systemprompt.io is built and delivered by the engineer who wrote it, with two years of hands-on AI governance experience. The same person who built the system trains your people and embeds it into their work.
One author, one product, full ownership. The software, the training, and the delivery all come from the person who built it. No hand-off to a junior team, no outside investors steering the roadmap. And if we disappear tomorrow, you keep the binary, the source, and the logs. That is the point of owning it.
A 30-minute call to scope what you need. We can implement it for you, or you can run it yourself. No prior setup or trial required. Prefer to try it first? Clone the template.
A focused 30-minute call. No preparation or prior evaluation needed.
We've sent you an email with a link to book your 30-minute call. Check your inbox.
While you wait: how to detect and govern shadow AI in your organisationOne team delivers the system, the training, and the rollout, then hands you the keys. Book a call to scope your engagement.